FFmpeg: libavformat/wv.c Source File

00001 /*
00002  * WavPack demuxer
00003  * Copyright (c) 2006,2011 Konstantin Shishkov
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include "libavutil/audioconvert.h"
00023 #include "libavutil/intreadwrite.h"
00024 #include "libavutil/dict.h"
00025 #include "avformat.h"
00026 #include "internal.h"
00027 #include "apetag.h"
00028 #include "id3v1.h"
00029 
00030 // specs say that maximum block size is 1Mb
00031 #define WV_BLOCK_LIMIT 1047576
00032 
00033 #define WV_EXTRA_SIZE 12
00034 
00035 #define WV_START_BLOCK  0x0800
00036 #define WV_END_BLOCK    0x1000
00037 #define WV_SINGLE_BLOCK (WV_START_BLOCK | WV_END_BLOCK)
00038 
00039 enum WV_FLAGS{
00040     WV_MONO   = 0x0004,
00041     WV_HYBRID = 0x0008,
00042     WV_JOINT  = 0x0010,
00043     WV_CROSSD = 0x0020,
00044     WV_HSHAPE = 0x0040,
00045     WV_FLOAT  = 0x0080,
00046     WV_INT32  = 0x0100,
00047     WV_HBR    = 0x0200,
00048     WV_HBAL   = 0x0400,
00049     WV_MCINIT = 0x0800,
00050     WV_MCEND  = 0x1000,
00051 };
00052 
00053 static const int wv_rates[16] = {
00054      6000,  8000,  9600, 11025, 12000, 16000, 22050, 24000,
00055     32000, 44100, 48000, 64000, 88200, 96000, 192000, -1
00056 };
00057 
00058 typedef struct{
00059     uint32_t blksize, flags;
00060     int rate, chan, bpp;
00061     uint32_t chmask;
00062     uint32_t samples, soff;
00063     int multichannel;
00064     int block_parsed;
00065     uint8_t extra[WV_EXTRA_SIZE];
00066     int64_t pos;
00067 }WVContext;
00068 
00069 static int wv_probe(AVProbeData *p)
00070 {
00071     /* check file header */
00072     if (p->buf_size <= 32)
00073         return 0;
00074     if (p->buf[0] == 'w' && p->buf[1] == 'v' &&
00075         p->buf[2] == 'p' && p->buf[3] == 'k')
00076         return AVPROBE_SCORE_MAX;
00077     else
00078         return 0;
00079 }
00080 
00081 static int wv_read_block_header(AVFormatContext *ctx, AVIOContext *pb, int append)
00082 {
00083     WVContext *wc = ctx->priv_data;
00084     uint32_t tag, ver;
00085     int size;
00086     int rate, bpp, chan;
00087     uint32_t chmask;
00088 
00089     wc->pos = avio_tell(pb);
00090     if(!append){
00091         tag = avio_rl32(pb);
00092         if (tag != MKTAG('w', 'v', 'p', 'k'))
00093             return -1;
00094         size = avio_rl32(pb);
00095         if(size < 24 || size > WV_BLOCK_LIMIT){
00096             av_log(ctx, AV_LOG_ERROR, "Incorrect block size %i\n", size);
00097             return -1;
00098         }
00099         wc->blksize = size;
00100         ver = avio_rl16(pb);
00101         if(ver < 0x402 || ver > 0x410){
00102             av_log(ctx, AV_LOG_ERROR, "Unsupported version %03X\n", ver);
00103             return -1;
00104         }
00105         avio_r8(pb); // track no
00106         avio_r8(pb); // track sub index
00107         wc->samples = avio_rl32(pb); // total samples in file
00108         wc->soff = avio_rl32(pb); // offset in samples of current block
00109         avio_read(pb, wc->extra, WV_EXTRA_SIZE);
00110     }else{
00111         size = wc->blksize;
00112     }
00113     wc->flags = AV_RL32(wc->extra + 4);
00114     // blocks with zero samples don't contain actual audio information and should be ignored
00115     if (!AV_RN32(wc->extra))
00116         return 0;
00117     //parse flags
00118     bpp = ((wc->flags & 3) + 1) << 3;
00119     chan = 1 + !(wc->flags & WV_MONO);
00120     chmask = wc->flags & WV_MONO ? AV_CH_LAYOUT_MONO : AV_CH_LAYOUT_STEREO;
00121     rate = wv_rates[(wc->flags >> 23) & 0xF];
00122     wc->multichannel = !!((wc->flags & WV_SINGLE_BLOCK) != WV_SINGLE_BLOCK);
00123     if(wc->multichannel){
00124         chan = wc->chan;
00125         chmask = wc->chmask;
00126     }
00127     if((rate == -1 || !chan) && !wc->block_parsed){
00128         int64_t block_end = avio_tell(pb) + wc->blksize - 24;
00129         if(!pb->seekable){
00130             av_log(ctx, AV_LOG_ERROR, "Cannot determine additional parameters\n");
00131             return -1;
00132         }
00133         while(avio_tell(pb) < block_end){
00134             int id, size;
00135             id = avio_r8(pb);
00136             size = (id & 0x80) ? avio_rl24(pb) : avio_r8(pb);
00137             size <<= 1;
00138             if(id&0x40)
00139                 size--;
00140             switch(id&0x3F){
00141             case 0xD:
00142                 if(size <= 1){
00143                     av_log(ctx, AV_LOG_ERROR, "Insufficient channel information\n");
00144                     return -1;
00145                 }
00146                 chan = avio_r8(pb);
00147                 switch(size - 2){
00148                 case 0:
00149                     chmask = avio_r8(pb);
00150                     break;
00151                 case 1:
00152                     chmask = avio_rl16(pb);
00153                     break;
00154                 case 2:
00155                     chmask = avio_rl24(pb);
00156                     break;
00157                 case 3:
00158                     chmask = avio_rl32(pb);
00159                     break;
00160                 case 5:
00161                     avio_skip(pb, 1);
00162                     chan |= (avio_r8(pb) & 0xF) << 8;
00163                     chmask = avio_rl24(pb);
00164                     break;
00165                 default:
00166                     av_log(ctx, AV_LOG_ERROR, "Invalid channel info size %d\n", size);
00167                     return -1;
00168                 }
00169                 break;
00170             case 0x27:
00171                 rate = avio_rl24(pb);
00172                 break;
00173             default:
00174                 avio_skip(pb, size);
00175             }
00176             if(id&0x40)
00177                 avio_skip(pb, 1);
00178         }
00179         if(rate == -1){
00180             av_log(ctx, AV_LOG_ERROR, "Cannot determine custom sampling rate\n");
00181             return -1;
00182         }
00183         avio_seek(pb, block_end - wc->blksize + 24, SEEK_SET);
00184     }
00185     if(!wc->bpp) wc->bpp = bpp;
00186     if(!wc->chan) wc->chan = chan;
00187     if(!wc->chmask) wc->chmask = chmask;
00188     if(!wc->rate) wc->rate = rate;
00189 
00190     if(wc->flags && bpp != wc->bpp){
00191         av_log(ctx, AV_LOG_ERROR, "Bits per sample differ, this block: %i, header block: %i\n", bpp, wc->bpp);
00192         return -1;
00193     }
00194     if(wc->flags && !wc->multichannel && chan != wc->chan){
00195         av_log(ctx, AV_LOG_ERROR, "Channels differ, this block: %i, header block: %i\n", chan, wc->chan);
00196         return -1;
00197     }
00198     if(wc->flags && rate != -1 && rate != wc->rate){
00199         av_log(ctx, AV_LOG_ERROR, "Sampling rate differ, this block: %i, header block: %i\n", rate, wc->rate);
00200         return -1;
00201     }
00202     wc->blksize = size - 24;
00203     return 0;
00204 }
00205 
00206 static int wv_read_header(AVFormatContext *s,
00207                           AVFormatParameters *ap)
00208 {
00209     AVIOContext *pb = s->pb;
00210     WVContext *wc = s->priv_data;
00211     AVStream *st;
00212 
00213     wc->block_parsed = 0;
00214     for(;;){
00215         if(wv_read_block_header(s, pb, 0) < 0)
00216             return -1;
00217         if(!AV_RN32(wc->extra))
00218             avio_skip(pb, wc->blksize - 24);
00219         else
00220             break;
00221     }
00222 
00223     /* now we are ready: build format streams */
00224     st = avformat_new_stream(s, NULL);
00225     if (!st)
00226         return -1;
00227     st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
00228     st->codec->codec_id = CODEC_ID_WAVPACK;
00229     st->codec->channels = wc->chan;
00230     st->codec->channel_layout = wc->chmask;
00231     st->codec->sample_rate = wc->rate;
00232     st->codec->bits_per_coded_sample = wc->bpp;
00233     avpriv_set_pts_info(st, 64, 1, wc->rate);
00234     st->start_time = 0;
00235     st->duration = wc->samples;
00236 
00237     if(s->pb->seekable) {
00238         int64_t cur = avio_tell(s->pb);
00239         ff_ape_parse_tag(s);
00240         if(!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
00241             ff_id3v1_read(s);
00242         avio_seek(s->pb, cur, SEEK_SET);
00243     }
00244 
00245     return 0;
00246 }
00247 
00248 static int wv_read_packet(AVFormatContext *s,
00249                           AVPacket *pkt)
00250 {
00251     WVContext *wc = s->priv_data;
00252     int ret;
00253     int size, ver, off;
00254     int64_t pos;
00255 
00256     if (url_feof(s->pb))
00257         return AVERROR(EIO);
00258     if(wc->block_parsed){
00259         if(wv_read_block_header(s, s->pb, 0) < 0)
00260             return -1;
00261     }
00262 
00263     pos = wc->pos;
00264     off = wc->multichannel ? 4 : 0;
00265     if(av_new_packet(pkt, wc->blksize + WV_EXTRA_SIZE + off) < 0)
00266         return AVERROR(ENOMEM);
00267     if(wc->multichannel)
00268         AV_WL32(pkt->data, wc->blksize + WV_EXTRA_SIZE + 12);
00269     memcpy(pkt->data + off, wc->extra, WV_EXTRA_SIZE);
00270     ret = avio_read(s->pb, pkt->data + WV_EXTRA_SIZE + off, wc->blksize);
00271     if(ret != wc->blksize){
00272         av_free_packet(pkt);
00273         return AVERROR(EIO);
00274     }
00275     while(!(wc->flags & WV_END_BLOCK)){
00276         if(avio_rl32(s->pb) != MKTAG('w', 'v', 'p', 'k')){
00277             av_free_packet(pkt);
00278             return -1;
00279         }
00280         if((ret = av_append_packet(s->pb, pkt, 4)) < 0){
00281             av_free_packet(pkt);
00282             return ret;
00283         }
00284         size = AV_RL32(pkt->data + pkt->size - 4);
00285         if(size < 24 || size > WV_BLOCK_LIMIT){
00286             av_free_packet(pkt);
00287             av_log(s, AV_LOG_ERROR, "Incorrect block size %d\n", size);
00288             return -1;
00289         }
00290         wc->blksize = size;
00291         ver = avio_rl16(s->pb);
00292         if(ver < 0x402 || ver > 0x410){
00293             av_free_packet(pkt);
00294             av_log(s, AV_LOG_ERROR, "Unsupported version %03X\n", ver);
00295             return -1;
00296         }
00297         avio_r8(s->pb); // track no
00298         avio_r8(s->pb); // track sub index
00299         wc->samples = avio_rl32(s->pb); // total samples in file
00300         wc->soff = avio_rl32(s->pb); // offset in samples of current block
00301         if((ret = av_append_packet(s->pb, pkt, WV_EXTRA_SIZE)) < 0){
00302             av_free_packet(pkt);
00303             return ret;
00304         }
00305         memcpy(wc->extra, pkt->data + pkt->size - WV_EXTRA_SIZE, WV_EXTRA_SIZE);
00306 
00307         if(wv_read_block_header(s, s->pb, 1) < 0){
00308             av_free_packet(pkt);
00309             return -1;
00310         }
00311         ret = av_append_packet(s->pb, pkt, wc->blksize);
00312         if(ret < 0){
00313             av_free_packet(pkt);
00314             return ret;
00315         }
00316     }
00317     pkt->stream_index = 0;
00318     wc->block_parsed = 1;
00319     pkt->pts = wc->soff;
00320     av_add_index_entry(s->streams[0], pos, pkt->pts, 0, 0, AVINDEX_KEYFRAME);
00321     return 0;
00322 }
00323 
00324 static int wv_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int flags)
00325 {
00326     AVStream *st = s->streams[stream_index];
00327     WVContext *wc = s->priv_data;
00328     AVPacket pkt1, *pkt = &pkt1;
00329     int ret;
00330     int index = av_index_search_timestamp(st, timestamp, flags);
00331     int64_t pos, pts;
00332 
00333     /* if found, seek there */
00334     if (index >= 0 &&
00335         timestamp <= st->index_entries[st->nb_index_entries - 1].timestamp) {
00336         wc->block_parsed = 1;
00337         avio_seek(s->pb, st->index_entries[index].pos, SEEK_SET);
00338         return 0;
00339     }
00340     /* if timestamp is out of bounds, return error */
00341     if(timestamp < 0 || timestamp >= s->duration)
00342         return -1;
00343 
00344     pos = avio_tell(s->pb);
00345     do{
00346         ret = av_read_frame(s, pkt);
00347         if (ret < 0){
00348             avio_seek(s->pb, pos, SEEK_SET);
00349             return -1;
00350         }
00351         pts = pkt->pts;
00352         av_free_packet(pkt);
00353     }while(pts < timestamp);
00354     return 0;
00355 }
00356 
00357 AVInputFormat ff_wv_demuxer = {
00358     .name           = "wv",
00359     .long_name      = NULL_IF_CONFIG_SMALL("WavPack"),
00360     .priv_data_size = sizeof(WVContext),
00361     .read_probe     = wv_probe,
00362     .read_header    = wv_read_header,
00363     .read_packet    = wv_read_packet,
00364     .read_seek      = wv_read_seek,
00365 };